Drawer
The Drawer is a persistent side panel that provides live feedback during your chat. It shows what DeepLore Enhanced is doing in real time: which entries were injected, why, and how your token budget is being used.
The Drawer opens automatically when DLE is enabled and has data to display.
The Drawer has three zones:
- Status Zone (fixed at top) — connection status, stats, token/entry bars, quick actions
- Tabs (scrollable middle) — four tabs with different views
- Footer (fixed at bottom) — context window bar, health indicators, AI stats
- Status dot — color-coded: green (connected, healthy), yellow (indexing/stale), red (disconnected/error)
- Pipeline label — current search mode (Two-Stage, AI Only, Keywords Only)
- Stats row — entry count and pipeline mode at a glance
- Token budget bar — how much of your configured token budget is being used. Color gradient: green (low usage) through yellow to red (near/at budget)
- Entry count bar — entries injected vs. configured max entries cap
- Active gating filters — shows currently set era, location, scene type, and present characters (if any)
- Quick actions — Refresh Index, Open Settings, Run Scribe, Suggest New Lore
Shows what happened in the last generation — which entries were injected and which were filtered out.
Filter toggle: Switch between viewing Injected entries, Filtered entries, or Both.
Injected entries show:
- Entry title with match label (CONST, PIN, INIT, SEED, KEY, AI)
- For AI matches: confidence level and selection reason
- For keyword matches: the triggering keyword(s)
- Token count with color-gradient indicator
- Temperature indicator (see Temperature Heatmap below)
- Browse navigation button (arrow icon) — jumps to this entry in the Browse tab
Filtered entries show categorized rejection reasons:
- Budget/limit cuts
- Contextual gating filters
- Requires/excludes gating
- Cooldown, warmup, probability
- Re-injection cooldown
- Strip dedup
- AI rejection
- Click an entry to expand its full diagnosis
Generation diff: When the injected set changes between generations, badges show +N new and -N removed with brief reasons for removals (e.g., "Bootstrap fall-off", "No longer matched").
Searchable, filterable list of all indexed entries with virtual scrolling for large vaults.
Search: Type to filter by title or keyword (300ms debounce).
Status filters (8 options):
- All — show everything
- Injected — entries injected in the last generation
- Pinned — entries pinned to this chat
- Blocked — entries blocked from this chat
- Constant — always-inject entries (
#lorebook-always) - Seed — seed entries (
#lorebook-seed) - Bootstrap — bootstrap entries (
#lorebook-bootstrap) - Never Injected — entries that have never been injected (all-time analytics)
Tag filter: Dynamic dropdown populated from all tags found in the vault.
Custom field filters: Dropdowns for any custom gating fields defined in DeepLore/field-definitions.yaml are automatically added alongside the tag filter.
Sort options (7 options):
- Priority ascending/descending
- Alphabetical A-Z / Z-A
- Token count ascending/descending
- Injection count descending (most-injected first, this chat)
Entry rows show:
- Title, priority badge, token count
- Per-chat injection count badge (e.g., "3x")
- Temperature heatmap coloring (see below)
- Status indicators (injected/pinned/blocked markers)
Expand an entry to see:
- Content preview (first 200 characters)
- Keywords list
- Pin/Block toggle buttons (per-chat)
- Deep link to open in Obsidian
View and edit the current contextual gating filters for this chat, including both built-in and custom fields.
- Era — current era filter (with entry impact count: "filtering N entries")
- Location — current location filter
- Scene Type — current scene type filter
- Characters Present — list of characters currently flagged as present
- Custom fields — any user-defined gating fields also appear here with status dots and impact counts
Each filter has an Edit button that triggers the corresponding /dle-set-field command with its browse popup.
Manage Fields button (in the Gating tab toolbar) opens the rule builder UI where you can add, remove, or modify custom gating fields. Field definitions are stored in DeepLore/field-definitions.yaml in your vault. Each field has a type (text, number, boolean, list), a gating operator (equals, contains, any_of, none_of), and a tolerance level.
Entry Timers section shows active per-entry state:
- Cooldown entries with generations remaining
- Stale entries (above decay boost threshold, not recently injected)
- Decay penalties (injected too many consecutive times)
Quick-access buttons for all DLE slash commands, organized in four groups:
| Group | Tools |
|---|---|
| Diagnostics | Health Check, Inspect, Status, Simulate |
| AI Tools | AI Review, Notebook, Summarize |
| Data | Import World Info, Optimize Keys, Graph, Scribe History |
| Setup | Setup Wizard, Pins/Blocks, Help |
Each button triggers the corresponding /dle-* slash command. See Slash Commands for details.
- Context window bar — shows total context tokens used (prompt + lore + system) and reserved response tokens. Two-color bar: used (left) and reserved (right).
-
Health indicators — five icons with one-click access:
- Vault health (runs
/dle-health) - Connection status (runs
/dle-status) - Pipeline trace (runs
/dle-inspect) - Cache info popup
- AI service status (circuit breaker state, session stats)
- Vault health (runs
- AI session statistics — calls made, cache hits, tokens used (session-scoped, accumulates across chat switches)
The Browse tab uses color tinting to show injection frequency relative to the vault average:
- Hot entries (red tint) — injected more frequently than average. The more intense the red, the more over-represented this entry is.
- Cold entries (blue tint) — injected less frequently than average, or never injected.
- Neutral entries — near the average injection rate, no tinting.
Temperature is computed per-chat from injection counts. Constants and contextually-gated entries are excluded from the average calculation.
- Auto-opens when DLE is enabled and has entries to display
- Pin toggle — pin the drawer open so it stays visible across chat switches
- Overlay mode — on narrow screens (chat width > 60% of viewport), the drawer switches to an overlay that can be dismissed
- Tab persistence — remembers your last-viewed tab across sessions
- Live updates — refreshes automatically after each generation via observer callbacks (no manual refresh needed)
- Accessibility — ARIA roving tabindex, keyboard navigation, screen reader announcements via live region, reduced-motion support